Sliding DFT Tutorial (1 of 6)

Nearly all signal processing systems that use algorithms such as the fast Fourier transform (FFT) process signal samples by grouping them into batches. The starting points of any two adjacent batches are several tens or even thousands of samples apart in a typical system. In other words, the processor jumps through the data, taking snapshots of the data at regular intervals. The effect on the output of the signal processor is similar to using a strobe lamp, rather than a continuous light source, to keep track of the signal's behavior.

The difference between the time that an interesting event occurs and the time that the processor generates a detection report can be as long as one batch period. This latency can cause difficulties in time-critical systems.

If the event occurs at the boundary between two batches, it can prevent the processor from producing a reliable detection report.